MIL- STD-883F 2004 TEST METHOD STANDARD MICROCIRCUITS - 第604页

MIL-STD-883F METHOD 5004.11 18 June 2004 14 APPENDIX A g. Notc hing h. Tunneli ng i. Cusp ing 40.4 Defec t char acter izati on must i dentif y and quantify non- cr itic al defec ts , cr itic al defec ts , and ki ller def…

100%1 / 708
MIL-STD-883F
METHOD 5004.11
18 June 2004
13
APPENDIX A
- Diffusion
Pattern
Violation: Bridging between wells, width reduction (resistors) and enlargement.
- Dielectric
Film
Faults: Blown contacts/via's, holes, cracking, active junction line exposure,
excessive thickness variations.
- Die Surface
Protection
Faults: Cracks, pinholes, scratches, voids, cornerholes, peeling/lifting, blistering,
bond pad clearance.
- Diffusion,
isolation
defects,
trenches,
guard rings,
other
techniques: Voids, notches in pattern diffusion, overlaps of diffusion, contact windows
tub-to-tub connections (except by design), etc.
DEFECTS
: EXAMPLES/TYPES/CONSIDERATIONS:
- Film
Resistor
Flaws: Scratches, voids, potential bridging, non-adherence, corrosion, alignment,
overlap between resistors and conductive traces, step coverage thinning,
composition (color) changes.
- Laser Trimmed
Film Resistor
Flaws: Kerf width, detritus, current carrying violations (resistor width).
- Foreign
Material: Foreign to process step/ structure (chemical stains, photoresist, ink, stains,
liquid droplets).
Note: See appropriate category figures in TM 2010 Conditions A and B
40.3.1 The following metallization concerns need to be addressed by the manufacturer in the process control procedures
used to demonstrate metal integrity.
a. Silicon consumption
b. Junction spiking
c. Silicon precipitates (nodules)
d. Copper nucleation
e. Nonplanarity
f. Undercutting
MIL-STD-883F
METHOD 5004.11
18 June 2004
14
APPENDIX A
g. Notching
h. Tunneling
i. Cusping
40.4 Defect characterization must identify and quantify non-critical defects, critical defects, and killer defects at each
mask level and establish action limits at the appropriate inspection steps. If 100% in-line or end-of-line production screens
are used to remove a specific defect, action limits and inspections may not be required at the affected mask level.
Characterization must determine the major sources of variations and the impact of defect attributes (ie: size, mass,
composition and quantity). Characterization must comprehend the effects of defects on the mask level being characterized
and their impact on subsequent mask levels, up to and including the final product. Characterization must encompass defect
behavior at worse case allowable processing locations (eg: worse case physical location for critical defect generation), at
worse case boundary conditions (ie: thickness, temperature, gas flow, etc.) and to worse case design rules. See
Attachment #1: Example of Defect Characterization.
40.5 In accordance with the results of defect characterization, the action limits for defects must be less than the level at
which the defects are known to adversely affect the reliability and performance of the device (the use of process "safety
margins" must be invoked, eg: if an aluminum line with a 25% notch is known to shorten the life of the device, then margin
limits for the notching must be accounted for, that is, the allowable notch limit must be less than 25%). By definition, any
observation of a killer defect (one or more) exceeds its action limit.
40.6 The results of the defect characterization shall be used to establish inspection sampling requirements (ie: sample
sizes and sampling frequency) and analytical techniques for in-line and end-of-line process inspections (see section 50).
40.7 The manufacturer shall establish a process baseline and put the process under formal change control after defect
characterization has been completed and in-line and end-of-line inspection steps are implemented. Any changes that
adversely affect the defects require re-characterization of the defects (eg #1: changing fabrication gowns may affect
particulate generation and must be determined if they are equivalent or better than gowns used when the original defect
characterization was completed, if better no further action, if worse, re-characterization of the line. eg #2: a change in HCl
(hydrochloric acid) chemical supplier requires comparative analysis of new supplier to old supplier, relative to trace
impurities, followed by an engineering evaluation to validate the impact on product. Discovery of excessive, new impurities
that could not be proven benign would require re-characterization before the new supplier could be used).
40.8 Any manufacturers' imposed in-line or end-of-line screens must demonstrate their effectiveness in eliminating killer
and critical defects in excess of their allowable action limit(s).
40.9 Any new defects that surface as a result of excursion containment, yield analysis, customer returns, inspection
procedures, (etc.) must be characterized in accordance with specifications in section 40.
50. Inspection and test system
:
50.1 Control and reduction of defects will result from an inspection and test system, employing process and product
monitors and screens. The inspection and test system is incorporated throughout the wafer fab process flow (in-line and/or
end-of-line). It is expected that an inspection and test system will prevent killer defects from appearing in the delivered
product. See Attachment #2: Example of an Inspection and Test System.
50.2 Inspection and test procedures shall form an integrated approach that in total controls and reduces defects. The
procedure shall consider the following criteria where applicable:
MIL-STD-883F
METHOD 5004.11
18 June 2004
15
APPENDIX A
50.2.1 The supplier shall define and implement inspection and test procedures at appropriate points to monitor killer and
critical defects (as identified in section 40).
50.2.2 The inspection and test procedures shall consist of sampling plans which recognize the sources of defects and
their variance (ie: random, variation from die to die within a wafer, variation from wafer to wafer within a lot, variation from lot
to lot, variation with date of manufacturer). Sample plans shall be consistent with statistical practices (distributional form
and alpha/beta risks). The population to be sampled must be homogeneous.
Examples of homogeneity considerations include:
Lots that have been split or otherwise altered for rework are not considered homogeneous, unless otherwise
demonstrated, and therefore require independent sampling of the non-homogeneous (reworked) population. If
different pieces of processing equipment are used at the same process step (mask level), for the same purpose (eg:
use of multiple wafer steppers on the same wafer lot), these tools must demonstrate the killer and critical defect
characteristics are statistically comparable, for a given wafer lot to be considered a homogeneous population.
50.2.3 Inspections and tests must consider, but are not limited to worst case locations (as identified in Section 40).
Examples Include: 1) At an LPCVD operation, the defect characterization might determine particles to be consistently
higher on wafers at, or near, the door end of the tube, sampling at LPCVD must comprehend inspection at this location.
While characterizing metal bridging, one location on the die might appear consistently more prone to bridging than other die
locations, sampling criteria should include inspections at this location.
50.2.4 Inspection and test procedures must make use of "look backs". A look back inspection examines the current
process step and one or more preceding process steps. This procedure allows for inspection/test of telescoping effects
(magnifies or enhances the defect) and/or defects decorated by subsequent processing. This technique allows for additional
opportunities to inspect/test for killer and critical defects in preceding layers.
Examples Include: 1) While inspecting field oxidation it is possible to look back at pattern definitions in previous
levels. 2) A defect is known to be more obvious after a subsequent LTO deposition (the defect size telescopes),
therefore an inspection at LTO could effectively look back at the previous operation which generate a defect.
50.2.5 Inspection and test procedures must define action limits and the appropriate data to be recorded. Data recording
shall recognize the need for wafer, lot, or product disposition and corrective action (eg: data may need to be classified by
machine number, tool, wafer lot, operator, etc.). These types of data and action limits are derived from the defect
characterization (as identified in section 40) and shall take into account relevant attributes of defects (ie: size, color, mass,
composition, density). Action limits shall comprehend safety margins (as specified in section 40).
50.2.6 As a result of defect characterization (see section 40), non-critical defects shall be monitored, unless the non-
critical defect has been proven not to have any influence on the finished product, regardless of incidence or defect density.
This is required to address situations when:
a. Non-critical defects may mask detection of killer and critical defects (eg: a change in color obscures visual
observation of a killer or critical defect).
b. A non-critical defect becomes critical as a result of increased defect density (eg: due to an increase of non-critical
defects, a chain is formed, creating a critical defect).
c. An inconsistency between the incidence of non-critical and critical defects, signaling a change in the process that
must be explained.